NOTE: for the latest version of MIDAS, please visit: https://github.com/ranjitlall/MIDAS
MIDAS draws on recent advances in deep learning to deliver a fast, scalable, and high-performance solution for multiply imputing missing data. MIDAS employs a class of unsupervised neural networks known as denoising autoencoders, which are capable of producing complex yet robust reconstructions of partially corrupted inputs. To enhance their efficiency and accuracy while preserving their robustness, these networks are trained with the recently developed technique of Monte Carlo dropout, which is mathematically equivalent to approximate Bayesian inference in deep Gaussian processes. Preliminary tests indicate that, in addition to handling larger datasets than existing multiple imputation algorithms, MIDAS generates more accurate and precise imputed values in ordinary applications.
To install MIDAS via pip, input the following command into the terminal:
pip install git+https://github.com/Oracen/MIDAS.git
MIDAS requires
Tensorflow also has a number of requirements, particularly if GPU acceleration is desired. See https://www.tensorflow.org/install/ for details.
Variational autoencoder enabled. More flexibility in model specification, although defaulting to a simple mirrored system. Deeper analysis tools within .overimpute() for checking fit on continuous values. Constructor code deconflicted. Individual output specification enabled for very large datasets.
Key added features
Planned features:
Wish list:
Alpha 0.1:
Basic functionality feature-complete.